Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
1
Direct drive by cyclotron heatingcan explain spontaneous rotation
in tokamaks
J. W. Van Dam and L.-J. Zheng
Institute for Fusion StudiesUniversity of Texas at Austin
12th US-EU Transport Task Force Annual MeetingSan Diego, CA – April 16-20, 2007
2
Significance of rotation
• Rotation and velocity shear areimportant in tokamaks for confinementand stabilization– Stabilize resistive wall modes– Suppress turbulent transport
• Neutral beams, which provide rotationin current-day tokamaks, may beinsufficient in reactor-grade plasmas(e.g., ITER)– Short penetration depth– Requires high injection energy E, hence
modest imparted momentum ~ Pinj / E1/2
• Hence, considerable interest in theintrinsic toroidal rotation observed to bespontaneously generated, withoutexternally applied torque, in tokamaksheated with waves in the cyclotronfrequency range
Toroidal rotation with ICRHin C-Mod (Rice et al. 2001)
ΔvTo
r (10
4 m
/s)
ΔW/Ip (kJ/MA)
3
Experimental observations
• Remarkably similar results for intrinsictoroidal rotation from devices with differentheating methods and plasma shapes andconditions– ICRH in JET, Alcator C-Mod, and Tore
Supra: rotation in co-current direction for H-mode or other advanced confinement regimeplasmas
– ECH in DIII-D and TCV: counter-current– LH and ECH in JT-60U
• Experiments find intrinsic rotation velocity isproportional to plasma stored energy (orpressure) and scales inversely with current
• Intrinsic rotation is an appreciable fraction ofAlfvén speed– Possibly strong enough to stabilize MHD
modes in ITER
Toroidal rotation with ICRH andECH (Rice et al., IAEA 2006)
4
Earlier theories
• Various theories have been proposed to explain the generation of intrinsicrotation without auxiliary momentum injection, based on:– Neoclassical or sub-neoclassical effects [Kim (1991), Rogister (2002)]– Radial orbit shifts of cyclotron-heated energetic ions [Chang (1999), Chan (2002),
Eriksson (2004)]– Turbulence-induced toroidal stress [Shaing (2001)]– Electrostatic modes driven by the ion pressure gradient [Coppi (2002)]– Blob transport at the plasma edge [Myra (2006)]
• These theories have been carefully compared with experimental observationsrecently– Although each of these theories has its merits, the underlying mechanism for intrinsic
rotation is still uncertain– Rice (IAEA 2006): “At present there is no comprehensive, quantitative theoretical
explanation of spontaneous/intrinsic rotation.”
• New theory by Gurcan et al. (plenary talk at this meeting)– Non-diffusive Reynolds stress and sheared ExB symmetry breaking
5
New theory
• We propose an alternative explanation, showing that cyclotron waveheating can provide a direct drive for intrinsic toroidal rotation of thecore plasma
• Even though cyclotron heating is applied with a symmetric spectrum ofwaves, without a preferred toroidal direction, we find that, when theeffects of finite orbit size and magnetic field inhomogeneity are takeninto account, the toroidal momentum input due to cyclotron waveheating is actually unbalanced in the toroidal direction, thus causingrotation
• We propose that this type of direct rotation drive could provide anexplanation for the spontaneous rotation due to cyclotron heating
6
Difference with other fast-ion theories
• Earlier fast-ion theories consideredthe radial electric field (or radialcurrent) produced by a radiallyoutward or inward shift of trappedfast ions as the cause for rotation
Poloidal projection of guiding centerorbit for trapped hydrogen ion withon-axis ICRF perpendicular heating(Chang et al. 1999)
• Canonical momentum pφ is aninvariant unless wave-particleresonance occurs.
p! = mRv! " e#
• With the approximation (near thebanana tips, at resonance plane)
v! " v||# 0
an increase in pφ leads to radialoutward motion
• But this neglects precessionalmotion
7
Our theory includes precessional drift motion
• Our theory takes into accountthe toroidal precessionalmotion of the trapped ions– Since the toroidal
precessional drift has aspecific direction, trappedions resonantly interact onlywith an asymmetric portion ofthe cyclotron wave spectrum
– By absorbing energy throughcyclotron heating, the rapidlyprecessing ions increase innumber; this causes thetoroidal rotation likewise toincrease
– The fast resonant ions cantransfer their momentum tothe bulk plasma throughcollisions
Trapped particle orbits in tokamaks
8
Theoretical approach
• We use quasi-linear theory in action-angle variables (J, θ) to solve forthe perturbed distribution function δf and the slowly diffusing zeroth-order averaged distribution function <F0>:
• For wave absorption, only trapped particles will be considered, sincethey have more time than passing particles to interact with the appliedwaves
9
Solution of quasi-linear equation
• We use quasilinear theory:
• We use quasilinear theory:
Toroidal precessionfrequency (time-averaged)
10
Calculation of ICRF-induced torque
• We derive the rate of total momentum input (mechanical) due tocyclotron heating:
• Here, Pw is the energy absorption rate per volume from the RF waves:
• Also, nres is the density of resonant trapped particles:
(with + sign for ICRH and - sign for ECH)
– We take κ = sin (θt/2) ~ constant, since only trapped particles at the banana tips canhave significant absorption of ICRH energy and momentum
• This result seems to explain a number of the experimental observations
11
Scaling of toroidal rotation velocity
• Convert the rate of total momentum input (torque T) due to cyclotronheating, to toroidal rotation velocity vφ:
– Use Tζ = n m (dvφ/dt) and Pw = n d(ΔW)/dt, where ΔW = stored energy(due to ICRF heating).
– Also, recall that the safety factor q is inversely proportional to plasmacurrent Ip: specifically, Ip = 2π a2 BT / R q(a) in the large aspect ratio limit.
• Derive a formula for the (incremental) toroidal rotation velocity vφ:
v! ="eD(# )
a
R
$%&
'()
*WIp
$
%&
'
()
– This reproduces experimental scaling with energy and plasma current
12
Scaling with temperature ratio
• Formula for the (incremental) toroidal rotation velocity vφ:
v! ="eD(# )
a
R
$%&
'()
*WIp
$
%&
'
()
• Experimentally, DeGrassie et al. (APS/DPP 2006) found that theintrinsic toroidal rotation velocity for the case of ECH, if multiplied by theratio of central temperatures Ti(0) / Te(0), provides a better comparisonwith rotation data from ICRH discharges.
• The theory displays this feature, since the stored energy (or powerabsorbed from the cyclotron waves) is proportional to the centraltemperature:
!We
=Te(0)
Ti(0)
"
# $
%
& ' !We
13
Direction of toroidal rotation velocity
∝ (e Ip)-1
Theoretical vφ
Direction:
– Co-current for ICRF– Counter-current for ECH(i.e., reversed in centralregion of ECH deposition)– Reverses for either ICRH orECH when direction of thecurrent is reversed
Experimentallyobserved rotation
• Consistent with experimental observations:
• Theoretical formula for the (incremental)toroidal rotation velocity vφ:
v! ="eD(# )
a
R
$%&
'()
*WIp
$
%&
'
()
Toroidal rotation profile inDIII-D, with ECH powerdeposition profile indicated(deGrassie 2004)
14
Rotation reversal for off-axis ICRH
• Theoretical formula for the(incremental) toroidal rotationvelocity vφ:
v! ="eD(# )
a
R
$%&
'()
*WIp
$
%&
'
()
Drift reversal due tolarge θt at small r
Theoretical vφ
Rotation reversal:Near-axis rotation is reversedduring off-axis heating in ITBdischarge
Experimentallyobserved rotation
Rotation profile during off-axis heating (Rice 2004)
15
Inverted radial profile evolution
• Experimental observations show:– In initial stages, the intrinsic toroidal rotation has a
radially inverted profile, being stronger at large radiiand weaker toward the center
– After transition from L-mode to H-mode the rotationand stored energy in the core rapidly increase
• In other theories this feature is interpreted tomean that rotation is transported from outerregion to center
• Our theory provides a different explanation– The function D(κ) depends on magnetic shear s,
through the banana width effect. For monotonicallyincreasing q(r) profile, the magnitude of Dincreases with minor radius r and could thuscontribute to a radially increasing rotation profile
– Also, following an L-to-H transition, the centralplasma density is known to become peaked; thiscould lead to enhanced absorption of wave energyin the center, causing central rotation to beincreased
D(κ) vs r , with shear s asparameter: --- off-axis heating -.-.- on-axis heating
v! ="eD(# )
a
R
$%&
'()
*WIp
$
%&
'
()
16
Predicted magnitude of rotation drive
• Energy confinement time is shorter than the momentumconfinement time; hence, use the linear stage (betweenred lines) to estimate the ratio of rotation velocity toenergy gain
• Experiment: Δvφmax [m/s] / ΔW [J] = 0.3
• Theory: Δvφ [m/s] / ΔW [J] = 0.1 fv <q>v / A– fv ~ 2-3 is ratio of central rotation speed to volume-
averaged rotation speed– <q>v ~ 2-3 is volume-averaged safety factor– A = 2 is ratio of ion mass to proton mass
THUS: Δvφ [m/s] / ΔW [J] = 0.2–0.5
• If normalize to plasma current:– Theory givesΔvφ [m/s] / {ΔW [J] / Ip [A]} = 3 x 105 (τmom/τE)
– Experiment findsΔvφ [m/s] / {ΔW [J] / Ip [A]} = 106
Linear phase
Time evolution of the storedenergy, rotation, etc. (Rice 1998)
17
Scaling in dimensionless variables
• Can incorporate machine sizeinformation by recasting therotation scaling in terms ofdimensionless variables– Ion thermal or Alfvén Mach
number Mi, A = vφ / vi, A– Normalized beta βN = β / (Ip/aBT)
• Data from different machines thencoincide (Rice et al., IAEA 2006)
• Our theory is easily rewritten indimensionless variables:
MA / βN = π m3/2/eRn1/2
Ion thermalMach number
Alfvén Machnumber
18
Intrinsic rotation in Ohmic plasmas
• Interestingly, intrinsic rotation hasalso been observed in Ohmicallyheated plasmas, without anycyclotron wave heating– Scaling of Ohmic rotation with stored
energy and plasma current is sameas for cyclotron wave heating,although the magnitude is smaller;also the flow is mostly in the SOLregion
– Our theory apparently does notprovide an explanation for this. Arethere multiple mechanisms?
• Recent work by Aydemir [APS/DPP2006] does seem to explain it– Self-consistent flows generated by
transport in non-ideal MHDequilibrium (with resistivity, bootstrapcurrent, neoclassical effects)
H-mode rotation velocities,with and without RF heating(Hutchinson et al. 2000)
19
Dipole equilibrium flow pattern ina double-null (DN) configuration
A. Aydemir [APS/DPP 2006]
• A non-uniform “classical resistivity”profile with S0=106, and SSOL=102.
• A simple bootstrap current model,localized to the pedestal region, withJBS/J0=0.3 is used
• Toroidal beta = 5x10-3.
• The resulting flow has an AlfvénMach number M~10-2,corresponding to velocities of order104 m/s.
• Induced toroidal flow is also dipolarand has no net toroidal angularmomentum.
20
Equilibrium flows in single-null (SN)configurations
Interaction of the toroidal flow with the X-point transfers momentum to the vesselthrough the open field lines, leaving a net toroidal angular momentum contributionto the plasma. A. Aydemir [APS/DPP 2006]
21
Comparison with experimentally observedSOL flows
LaBombard et al.,Nucl. Fusion (2004)
A. Aydemir [APS/DPP 2006]
• Poloidal flow directions and locationsappear to be in agreement withexperimental observations
• When corrected for different sign oftoroidal flux used in numericalcalculations, contribution to core rotationis also in agreement
22
Summary
• Proposed a new explanation for intrinsic/spontaneous rotation ofcore plasma during cyclotron wave heating of tokamaks– Based on asymmetric precessional acceleration of trapped ions by ICRH
• Find agreement of theoretical predictions with key experimentalfeatures, such as:– Scaling of toroidal rotation velocity– Direction of rotation– Magnitude of rotation– Radial profile of rotation
• This may be one of several mechanisms for intrinsic rotation– Intrinsic rotation observed at plasma edge in purely Ohmically heated
discharges apparently requires another mechanism– For example, recent work by Aydemir [APS/DPP Annual Meeting 2006] is
able to explain the SOL rotation (and its dependence on divertor topology)observed in Ohmic tokamak plasmas (e.g., C-Mod)